Using Artificial Intelligence for Polyp Size Estimation

This study is a technical report on the application of artificial intelligence (AI) in estimating polyp size during colonoscopy procedures. Conducted from May to June 2021, the research utilized a colon endoscopy phantom model where artificial polyps were created and placed. The study involved a survey of gastroenterologists who assessed polyp sizes from videos recorded during the experiment. The results indicated that the AI system, Argus, demonstrated a higher accuracy rate of 96% in size estimation compared to the gastroenterologists' median accuracy of 77%. Additionally, the gastroenterologists made incorrect surveillance recommendations based on polyp size significantly more often than the AI system, with 34 incorrect recommendations compared to none from Argus. The findings suggest that AI can enhance the accuracy of polyp size estimation, which is crucial for determining appropriate surveillance intervals in colonoscopy, potentially leading to cost savings in future studies.